This Week in European Tech: Europe's Palantir problem, AI sovereignty & rise of venture secondaries

28:07You're in experimentation mode and obviously personally and as a business. Is that fair? I'd say so. I mean, I think we benefit from obviously the industry that we're in and where we sit, that we kind of get to see what some of the very best firms are doing. And we're kind of trying to sort of take their lead somewhat. But I think also being honest with ourselves, I think we are still in the early innings of a massive technology phase shift that's still yet to find its sort of equilibrium. I think I was at the UKPC, the ex-BVCA conference, I think when was this beginning of the year? And I was doing a panel and there was maybe, what, two 250 VCP crowd.

28:56And I asked, I said, be honest with me, how many of you are using AI where it's actively moving the needle for your workflow? It's really, really making a difference to your workflow. Not experimentation, but really, really it's embedded and making a difference. And only a small smattering of the 250 people in the room put their hands up. So I think it's a fair proxy for most businesses in the experimentation phase where they're just trying to work stuff out. Do we build? Do we buy? How do we integrate? What goes where? How does it all fit together? And which departments get affected with what?

29:27And I think this rolls into the whole Uber CEO story talking about capping spend because Because I'm not sure how better Uber's business is because of this mega spend in AI in the last four months. What have they proven? What have they done faster or better or smarter? That wasn't in the article. It was just the sheer fact they're spending a shed ton of cash on all of these AI toolkits. Mark Andreessen, he talks about this as sort of modern day alchemy. you know the kind of historically alchemy was this idea of being able to turn lead to gold and you know people are looking for the philosopher's stone and the ability to do that instead of thinking look if we can turn lead into gold what an amazing thing that would be and of course many people tried and nobody ever succeeded and what we've succeeded in doing with ai is to turn as andreason says sand into intelligence because we can literally literally take you know create silicon, which is created out of sand, and then turn that into sort of thinking capability, which is magic if there ever was anything.

30:31And suddenly, not only have we created this magic, but it's available to everybody. We all have it, right? Even consumers can go online and sort of use Gemini for free. So suddenly we all have this magical power and everybody's rushing to find out, well, how can you actually use that to become mega productive? And of course there is, as you say, Dan and Matt, a lot of experimentation. And yeah, that is why we're seeing, I think, so many of these gyrations right now is everybody's rushing, everybody's experimenting. Personally, I think it's super exciting. Of course, it's good. We have some bean counters that are going to look at the bottom line and make sure there's return on investment.

31:09I have no doubt that that will be the outcome. How will they do that, Matt? You've got a great brain for this kind of thing. How would you measure ROI? Well, I mean, that's easy. You look at cash flows. At the end of the day, everything has to translate into money. That's why money is so great, right? Because it's the ultimate token. We can keep score of everything else. But people are sort of expecting that because they started using Claude and, I don't know, last quarter, it'll show up and massive profitability this quarter. Business doesn't move that quickly, guys. I mean, it takes a little bit of time to work out how everything should flow through and how to restructure and everything else.

31:44But it's happening. There's no doubt. Yeah, yeah. Well, let's talk about those that are making a shed ton of cash and are raising a shed ton of cash. What I've got next on the list is what I've called the$80 billion club. Everybody seems to be raising$80 billion. So Anthropic, Far To Go Public, jumping the queue of OpenAI. SpaceX, a near$2 trillion. I think their latest is$1.85. It moves a little bit every now and then. They're going live in a week. So they're going live on the 12th. It's the 5th of June as of recording. Alphabet, raising$85 billion on top of$100 billion in debt. Their little warranty, the 100-year bond.

32:21Warren Buffett, I know it's not him, but it is him. But is it? Is he still the chair? I don't know. Mads, I'm going to lean on you for that one. He is the patron saint of I Don't Get Tech. Even he's in for 10 bill into Alphabet. So the$80 billion club. Mads, tee this one up for us. We are all looking at these mecca IPOs that are lining up. It'll be the three biggest IPOs of all time, sort of all happening within one year. All very exciting for all market aficionados. The question some people are asking is, where's the money going to come from? Can the market absorb all these big IPOs? There's been a lot of analysis done on this.

32:59Goldman recently released a study saying, actually, the IPO activity right now is very low compared to what we've seen historically. This year, it looks like there'll be 100 IPOs, 250 in 2021. about 400 IPOs in 1999. So we're not at the top when it looks at sort of volume of IPOs. When we look at value, we've also seen issuance in terms of kind of total percentage of market cap as being below the historical norm. We've historically seen around 1.5 % of market cap being issued on average every year. Now it's around 1%. So on all these parameters we're looking at, It looks like the market will be able to absorb it just fine.

33:43And why is that? It's not necessarily because we need lots of new money to come into the market, but because corporates keep buying back their own stock. So buybacks are a huge, huge buyer. You have all the kind of the boring businesses over here buying back their own stock, and that's then flowing into hopefully great new technology companies that will make change in the world. And if that works the way we expect it will, Well, that is capitalism doing what it does best, taking money from where it's no longer needed and reallocating that to the next generation of good things. The floats are tiny, right, Mads?

34:17I mean, the floats are, I mean, 85 bill is obviously a lot of money. Anthropics is a little bit less than that, I think, but it's a lot of money. But in context, at a 1.75 trillion, wherever they land, the floats are actually, as a percentage, are actually quite small. Is that fair? That's exactly right. What Goldman's has also done is they've sort of tried to forecast that into the future and say, OK, initially you'll float 10 % of the company. What then happens over the following year as lockups start to expire? And even with those lockups coming off, the expectation is there'll still be plenty of demand in the market just because of the level of buybacks we're seeing.

34:56Well, at least, I mean, SpaceX. I mean, I've tried to make head and a tail of the S1, but I'm pretty sure that Elon and the core investors are locked in for over a year. I think it's like, whatever, 366 days. And that's what 50 % of the market cap. So I think, I don't know. I mean, there are a lot of lockup rules, pushes. I think six months is probably the median where there's going to be some more availability. Everything that we'll really know about these companies won't happen until 2027. Ooh, interesting. I mean, look, so just maybe double click on that one, Dan, because the saying from some of the big banks, they shall remain unnamed, is we always talk about conflicts of interest, and some of the more successful banks are saying, no conflict, no interest.

35:44And as we know, there are certainly some banks that are sitting on both sides of this, right? Goldman's there saying the market will be fine. They're also running the SpaceX IPO. So they're obviously well ensconced in what's going to happen here. And on their own analysis, they're sort of saying, look, the way we can get to a$1.78 trillion valuation is for the SpaceX AI revenue to increase by quite a lot over the next few years. They're saying from now on until 2030, it needs to increase by 100x, which is a fair chunk over four years. So sort of growing from$3 billion to sort of$300 billion. So is that possible?

36:33I mean, it'll all come down to space AI and whether people believe that that's going to happen in those timescales. Matt, if I could put you on the spot, who would you buy? we've we've always done well by having a portfolio in the in the very best companies we aren't we aren't we aren't stock pickers yeah for sure but i mean if you if you had your last hundred dollars who would you put it on are you an elon fan do you believe i mean i think you know i think people have always done well by backing you are you know like like clearly as i was about saying like there's a quite a large revenue bridge there right to get to those kind of numbers but like we've just seen time and time again that those top one percent you know or the what these are kind of top point one percent companies continuing to surprise to the upside like we've like we what we've always seen when we've underwritten secondary portfolios is we've always it's been too conservative on the winners time and time again.

37:37We've undermodeled it. And I think I'm pretty sure, is it David Clark, your CEO? I'm pretty sure I saw somewhere at some point the AI isn't a bubble video. And it feels like Vencap, and it was very comprehensive and thoughtful. And go to the Vencap site and look it up if you haven't seen it. But it was an interesting breakdown of why this isn't the bubble that it has been of years of old and decades past. So it's interesting that you say that you've always been a bit too conservative. Because, I mean, obviously there's a lot of real money and there's a lot of real activity that we haven't seen in other kind of bubblish, bubblish, bubbloos.

38:20What else did you want to go into, Mads? What else would you pick up on here? Other questions you asked about Alphabet, does Google really need the money? Is there any signal from Buffett in the Berkshire purchase, et cetera? I think Buffett is not making the investment decision here. I think that's Abel who's running the book. And he's really more of maybe an operator than a stock picker, if you could say. We don't have much of a view yet on exactly how he thinks about this, or at least he will say things, but we don't have sort of a decade long track record of saying this is how he's going to administer the money.

38:53So I think it's too early to read too much into that. Okay. I mean, obviously, Alphabet is all CapEx, right? This is all we need to build out. We've got such demand for our services. We need more data centers. We need more hardware provision. Is that right? Yeah. Yeah. Yeah. Okay. There was a story this week about America suing its own effectively. And as we know, the standard behavior is Europe legislates, America sues, and we kind of go around in this merry-go-round. But the table turned. America is suing its own. Florida just became the first state to take AI to court. Now, they're suing OpenAI and Sam Altman personally for an unsafe product of the back of a criminal probe tied to a campus shooting.

39:38The same week, Trump has watered down his AI order. I've forgotten the name of the order. If anyone remembers, please remind me. But it's basically made the whole AI regulation control piece largely meaningless. For an example, they had this 90-day mandatory review. It's now become a voluntary 30-day review, just as Mythos is now going to another 150 companies across 15 countries to be tested and trialled in the wild. So what is happening with Florida suing? Should they win? Will they win? Mads, can you tee this one up a bit better for us? I think it's an interesting development. I mean, kind of one might question why push it down that litigious route at all?

40:25And when does suing work? And when does it break? And why is Anthropik different from cold pizza? and I think it's just the recognition that there's some things that can break and then you can have a trial and you can kind of there can be remedies and that is okay in some situations and then there's some situations when it's not okay when it's kind of it's out of the bag we don't want you know somebody to take down the safety controls of a nuclear power plant and for there to be a big meltdown and then rely on kind of a civil court case afterwards then it's too late And so it's figuring out, you know, when is the harm bound to a sort of a more local issue that you can run a trial over?

41:09And when is it more unbound? Whereas once it's out of the bag, it's spreading and there's diffusion and it's just too late. And I think it's figuring out which is which. And to some extent, I think that's what we're seeing now is regulators and politicians are trying to work that piece out. Matt, have you got any thoughts on this one? Not much to add beyond that. I think it's that, yeah, some of those fundamental things around, like, what are those core guardrails, those kind of non-negotiables? And then inside of that, you know, I think we have a system that's quite well placed to iterate around errors.

41:47I wonder if we'll see any more. I wonder if we'll see, I wonder if this will open up other lawsuits or open up other challenges. It just feels like a weird, it just feels like a super weird thing for them to do. I mean, another vector in this, Dan, is kind of the naming, the personal naming of Sam Altman in one of the depositions, kind of one of the trials. The point there is, look, we have limitation of liability in our corporate codes for a good reason. It's sort of load bearing to entrepreneurship that you're able to stand up a business and the business is at risk, but people have a limitation of liability.

42:23So, you know, unless you really, if you do something that kills somebody, maybe you can have personal liability. But even there, I think the bar is quite high because, you know, at the end of the day, if business can't operate, if we can't innovate, well, then it's almost sort of back to the Stone Ages, right? So you need to have that limitation there. And yeah, again, I think there's something there to work out. I understand why people do it, but it feels maybe a little bit recessive. Yeah, and I guess we're pretty good at fining in the EU. I wonder if we'll see any other versions of this over here.

42:58I doubt it, but maybe we'll see more in the States. I have a prediction which is kind of aligned with what we were talking about before. So what we try and do, Matt, is I always look at the things that we've kind of researched and learned in the week. And then I try and work out what my prediction would be from all of the stuff that we've thought about and talked about. and is there something that I think that we'll take forwards? And I use it as a thought experiment. So I use it as like, you know, it's not necessarily a prediction set in stone. I'm going to put my children's lives on the lines here, but I use it as this kind of broader thought experiment.

43:32And here's mine for this week, which is that Europe cannot achieve total pure AI self-sufficiency. I don't think many would argue with that, but it will achieve strict sovereign control and will do it by splitting into two. So we've got the state tier, government critical infrastructure, will run solely in the EU on things like Mistral, Aleph Alpha in Germany, hosted on native cloud infrastructure, as we've talked about before. But the commercial tier will effectively run on tamed or legally compliant, bucketed and worked anthropic models, open AR models, run on this kind of hived off AWS or Azure infrastructure.

44:13And that's how we'll run the commercial tier. So So no great shakes there in my prediction, but I think we will in 27 see that two-tiering of AI, if you like, AI and hosting and infra. Anyone got any thoughts on that before we move on to deal of the week? Yeah, it seems plausible. I mean, I think there'll be lots of open source in use in many different situations because I just think there are many places where it makes sense. European, right? European allies? No, any kind of open source. Chinese? Yeah, why not? I can't see how we... My take also on this as a kind of a follow-on is that we're just not going to see...

44:53I've always thought that the Chinese OS projects would make much more inroads into corporate land through government infra, and I just can't see it anymore. I think I have to change my mind a little bit on this. I just can't see how it's going to pass any kind of regulatory muster. If you're doing anything in a regulated space or anything to do with touching data patient records or whatever it might be. But lots of workloads aren't regulated. Sure, sure, sure. Yes, yes. Obviously, in the light of private stuff, I'm with you. But anything meaningful, I just can't see Chinese OS making any inroads, which is a big shift because I always thought this would be a big part of the stack, but I'm not convinced anymore.

45:36Is that? Okay. I mean, just maybe you're right. I mean, just kind of the other side of the argument is that clearly coding is the most important workload right now, we're talking about important workloads. Coding is the most important. And there's a lot of Chinese open source software or open weight models that are used there, even if they are maybe slightly tweaked. We know that some of the stuff Cursor did before partnering with X was leveraging Chinese open source or open weight models as part of their coding stack. So I do think you'll see it, but maybe you're right. You may not be using it for patient records.

46:10i mean i've kind of put it in the mvp bucket as in we get off the tracks we start our own building our own models therefore we're going to use whatever is low cost and easy and and has the least restrictions at the lowest cost so i heard a i think just to just to jump in i had i i was i heard i heard a good analogy the other day thinking about like market construction around models and I was sort of like, like previously kind of had in my head that you'd end up in this sort of, I don't know, oligopoly equivalent to, I don't know, like the electricity industry or something. But like that, the error in that thinking is that the models are commoditized.

46:55Someone put forward quite a good analogy, which was thinking about them as airlines, as in a lot of upfront investment required, but they will all be quite different and serve different needs. And inevitably you'll end up with, yeah, so you're pouring down like those sort of low cost carriers. And then at the other end of the spectrum, you're going to have private charter working on with proprietary data sets at the super premium end of things. I thought that was quite a, I've not heard that before. It's quite a nice analogy. Yeah, no, I think that's a nice analogy. I also think one of my big previous predictions has been the on-prem piece, where we're going to start to see specific tooling purpose-built for within or for companies that live on-prem.

47:44And I think there's going to be a bit of a blended affair. We're going to start to get out of this experimentation phase and working out what we really need to run our businesses. And that will be a blend of different AI models delivered in different ways for different tasks within a business. And I think that's going to be really exciting as we see rubber hit road in those very specific ways to run the business. Mads, your deal of the week. Go. So last week we spoke about the Pope and we spoke about this. And we did. This kind of his new. It was a bit of a weird one, wasn't it? Getting all religious.

48:21His views on AI and how all that connects. And I think that sort of connects to, if not deal of the week, then heralded deal of the week in the sense of stock. They're talking about now raising 300 million euros at two and a half billion. I don't think it's closed yet, but it sounds like it's on track to happen. Headquartered in Berlin, stock is a defense technology company. And what do they do? They do the stuff the Pope has told them not to do, which is they have autonomous drones, kind of loitering munitions that'll sort of fly around and then ID a target and then strike. And they also have lots of AI.

49:00They have Minerva, which is sort of a command layer that can control things. And I just thought it was interesting. I don't know whether it's sort of, you know, the Catholics versus Protestant Germans or whatever. That's where the schism is. But all joking aside, I think we said at the time that there's absolutely no chance that we won't have autonomous weapons. Just can't see it because other sides will. We already have, I would assume. Yeah, to some extent, yes. Right. And the question is how far do you lean into it? And I just think that whether you like it or not, the move towards that is inexorable.

49:33It's going to come. And the question is then what do we do about it? Matt, is this the founder that's like 20, 20 or 22 years old? I saw a defense startup, two-year-old defense startup, raising, I think, just become a unicorn. And I'm pretty sure the founder's like 20, 21, 22, something like that. Is that this one? Whether he's just 21, I can't remember his age, Dan. I was reading, you've just stepped out or dropped out of uni and you've just started this defense tech business and you are an unbelievably young dude. Maybe it's not this one. I will find out and report back. My deal of the week is Gigaton.

50:14Now, this is the business that was formerly Carbon Re. Very nice business. I mean, boring to many. A love affair of ours, Mads. They help industrial companies run more efficient systems and release less carbon into the atmosphere. So they work with cement, steel, glass, chemicals firms. They've raised 26 mil to build out their control systems. So massive claps and a round of applause for Gigaton. I hope you absolutely knock it out of the park. What is happening this week? Matt, what are you up to this week? What's happening for Vencap or you or what's coming up in your diary that's exciting? Are you going on holiday?

50:50Are there any exciting deals on the table? What's happening for you? So, I mean, we're inundated with deal flow on the secondary side. so it all comes down to picking our spots it's a fascinating point in time just given the technology phase shift that we're going through, how we think about picking and pricing deals, I mean fortunately we can be choosy, which is great and yeah, hope to try and get a holiday in sometime, we've got an 8 month old baby at home that doesn't Is this number one? this is number one and he said he's not oh my goodness he doesn't sleep very well so uh oh he's got he's got four he's got four kids and i think he's been through every variant there is so you must you must speak with him about it i need i need some tips well i i had one that hate still he my son hates sleep he never slept as a baby he hates sleep and my daughter would sleep in a bomb site and um and so that was maybe i don't know if you're going to go again and have to two or three babies, but they're all different.

51:56So maybe hopefully the next one will sleep. Mads, any tips for sleeping? Yeah, I mean, earplugs are good, right? Yeah, we had hellish trouble with our boy. Mads, what is happening in the world and what is happening for you in the weekend? We've got a real damn day on Monday, don't we? I mean, it's kind of the Apple's Big WWDC. Bad boy, Matt, just to share the context there. Most of my predictions are wrapped around something that Apple will end up doing, which they invariably don't end up doing. But that's why Mads is raising that. So the question is, is the real Siri finally going to stand up?

52:37I bloody hope so. And is it going to be wrapped in Google or Anthropic Tech on the back end? All of it. I thought the new theory was going to be that you could add in any model that you wanted. It's going to be a bit of a marketplace. Fabulous. But that's not really the job style, is it? It's more like you can have any color you want as long as it's black. It's like we'll choose for you. I think this is a case of we really screwed the pooch. Let's just get the best and learn off the best. And then over time, we can build out our own. I assume. Interesting. I assume allegedly. So that's Monday, I believe.

53:12Thursday, we've got ECB meeting and it looks like there'll be a rate hike, which is not great for the old continent. obviously driven by the kind of Iran-driven energy shock and risk of stagflation and all that good stuff become on the back end. So not super keen on that. Friday, however, all the excitement returns. SpaceX should debut on the NASDAQ. It's a week to day. Biggest IPO of all time. It's finally happening. I'm assuming that it's going to be a bit of a nothing burger to start because everything is locked up and my assumption is this will do well out of the blocks, nothing to see here, but we'll really see how the market's going to react in, let's say between six and 12 months.

53:58Is that fair? I think it's going to, stuff's going to happen soon enough. Really? Really, really, really. Okay. Okay. Yeah. I mean, I think there's going to, yeah, I think Friday is going to be an exciting day in the public markets. I was just, I was just looking at poly market in terms of like odds of like closing market cap for Friday and there's 9 % that it's going to close at over 3 trillion. Really? Bananas. A 1 in 10 that it's going to close at 3. Yeah, over 3. It's the proper Elon magic, isn't it? Because nothing that I can see in the laws of physics and maths makes any sense underneath Elon magic.

54:42But as you said, don't ever bet against the man. but yeah okay no i mean the other thing just on as you said you know these these s ones are pretty dense the guys at meritech do really great jobs on um synthesizing s ones they've they've done one on spacex typically it's basically good around the lockup mechanism some of those um thresholds around performance of the stock and how that then translates to kind of what people can sell down So for a simpleton like me, it's quite useful. I'm definitely going to look that up because, I mean, I don't often, I mean, apart from for Upside, this pod, as VCs, we don't obviously tend to look at Publix and S1s.

55:29I mean, we do for the pod, but I mean, I'm really starting to lift my eyes off the desk and trying to understand more about some of the crazy things that go on elsewhere. So I would definitely look them up. All right. Well, thank you so much, my lovely gentleman. If there is nothing else, we will call it a day and we will see you all next week.